Solving Unconstrained Optimization Problems by a New Conjugate Gradient Method with Sufficient Descent Property

Abstract

There have been some conjugate gradient methods with strong convergence but numerical instability and conversely‎. Improving these methods is an interesting idea to produce new methods with both strong convergence and‎‏  ‎numerical stability‎. ‎In this paper‎, ‎a new hybrid conjugate gradient method is introduced based on the Fletcher ‎formula (CD) with strong convergence and the Liu and Storey formula (LS) with good numerical results‎. ‎New directions satisfy the sufficient descent property‎, ‎independent of line search‎. ‎Under some mild assumptions‎, ‎the global convergence of new hybrid method is proved‎. ‎Numerical results on unconstrained CUTEst test problems show that the new algorithm is ‎very robust and efficient‎.